Infant movement classification through pressure distribution analysis

Background Aiming at objective early detection of neuromotor disorders such as cerebral palsy, we propose a non-intrusive pressure sensing approach to classify infant general movements, differentiating typical patterns of the fidgety period (fidgety movements) vs. the pre-fidgety period (writhing movements). Methods Participants (N=45) were sampled from a typically-developing infant cohort. Multi-modal data, including pressure data from a 1024-sensor mat, were prospectively recorded across seven biweekly sessions from 4–16 weeks post-term age. From two targeted age periods, 1776 five-second pressure snippets were used for classification. Snippets were pre-annotated via synchronized video as fidgety present or absent. Support vector machines, feed-forward networks, convolutional neural networks, and long short-term memory networks were evaluated. Results The convolutional neural network achieved the highest average classification accuracy (balanced accuracy 81.4%). Compared to other automated general movement assessment methods, the pressure sensing approach shows high potential for clinical applications. Conclusions The pressure sensing approach offers efficient, non-intrusive large-scale motion data acquisition and sharing, with potential scalability for clinical evaluation of infant neuromotor functions.